Report

Thermoelectricity in Molecular Junctions

See allHide authors and affiliations

Science  16 Mar 2007:
Vol. 315, Issue 5818, pp. 1568-1571
DOI: 10.1126/science.1137149

Abstract

By trapping molecules between two gold electrodes with a temperature difference across them, the junction Seebeck coefficients of 1,4-benzenedithiol (BDT), 4,4′-dibenzenedithiol, and 4,4′′-tribenzenedithiol in contact with gold were measured at room temperature to be +8.7 ± 2.1 microvolts per kelvin (μV/K), +12.9 ± 2.2 μV/K, and +14.2 ± 3.2 μV/K, respectively (where the error is the full width half maximum of the statistical distributions). The positive sign unambiguously indicates p-type (hole) conduction in these heterojunctions, whereas the Au Fermi level position for Au-BDT-Au junctions was identified to be 1.2 eV above the highest occupied molecular orbital level of BDT. The ability to study thermoelectricity in molecular junctions provides the opportunity to address these fundamental unanswered questions about their electronic structure and to begin exploring molecular thermoelectric energy conversion.

Study of charge transport in molecules is of fundamental interest, with potential applications in molecular electronics (1) and energy-conversion devices (2, 3). Current-voltage (I-V) characteristics of single molecules have been extensively investigated by trapping a single molecule in break junctions formed by mechanical strain (4), electromigration (5), and scanning tunneling microscopes (6). Although such measurements have provided substantial insight into charge transport through molecular junctions, critical aspects about the electronic structure cannot be uniquely obtained by I-V characteristics alone. For example, whether molecular junctions are p-type or n-type—i.e., whether the position of the Fermi level, EF, of the metal contacts is closer to the highest occupied molecular orbital (HOMO) or to the lowest unoccupied molecular orbital (LUMO)—generally remains unknown because of uncertainties in the microscopic details of the contacts (710). A specific example of this is benzenedithiol (BDT); in this case, although the electrical conductance has been studied extensively both experimentally (4, 11) and theoretically (7, 1215), some groups suggest that the EF of the electrodes lies close to the HOMO level (7, 13, 15) and other groups contend that EF lies near the LUMO level (12, 14). Although sweeping the gate bias in single-molecule transistors could potentially yield this information, Coulombic interactions caused by charge accumulation on the molecule could perturb the electronic structure (16, 17).

It has been suggested that the sign of the Seebeck coefficient, S, of molecular junctions can indicate the sign of the charge carrier and the relative position of EF with respect to the HOMO or LUMO levels (8). Indeed, thermopower measurements that use a scanning probe microscope have yielded nanoscale spatial distributions of electron and hole concentrations in inorganic semiconductors (18) and have led to chemical potential microscopy at the atomic scale (19, 20). Here, we report an alternative approach: We measured S for molecular junctions formed by trapping molecules between gold electrodes, and then we measured the voltage generated across them when a temperature bias was imposed across the junction.

In general, S is associated with bulk materials and is obtained by measuring the voltage difference created across a material in response to an applied temperature differential. In such bulk materials, charge transport is diffusive in nature. The concept of an effective S is also valid for junctions where the transport may be ballistic. For such junctions, however, a more general form of the Seebeck coefficient is needed and is given as Embedded Image(1) where σ(E) is the energy-dependent differential electrical conductivity, EF is the Fermi level (or more accurately, the chemical potential), e is the charge of an electron, and T is the absolute temperature; the denominator in Eq. 1 is the electrical conductivity, σ. As Eq. 1 suggests, S reflects the asymmetry in σ(E) with respect to EF. In bulk materials, this asymmetry results from energy-dependent carrier scattering or the asymmetry in the density of states. For ballistic transport, the asymmetry can be created by a potential barrier at a junction, such as that created between EF of a metal and the HOMO or LUMO level of a molecule. Here, S is not an intrinsic property of a material, but that of the heterojunction. Hence, we call it a junction Seebeck coefficient, Sjunction. Because Sjunction measures the size of an energy barrier, it is not expected to depend on the number of molecules trapped between the electrodes and is, therefore, an intrinsic property of the junction. This is in contrast to the junction's electrical conductance, which depends on the number of molecules.

A modified scanning tunneling microscope (STM) setup is shown schematically in Fig. 1A: A customized control circuit drives a Au STM tip at a constant speed toward a Au substrate in air under ambient conditions. The Au tip is kept in contact with a large thermal reservoir at room temperature, which maintains the tip temperature very close to ambient (2123). The Au substrate can be heated with an electric heater to a desired temperature above ambient to create a tip-substrate temperature difference, ΔT. When the Au STM tip approaches the hot substrate, a tip-substrate voltage bias is applied and the current is continuously monitored. When the conductance reaches a sufficiently high threshold of 0.1Go, where Go = 2e2/h is the quantum of charge conductance [see (23) for a discussion on the choice of 0.1Go], our previous experiments on electrical conductance have shown that the tip-substrate distance is sufficient to trap molecules between the electrodes (24). Once this threshold is reached, the voltage bias and the current amplifier are disconnected, and the voltage amplifier is connected instead (Fig. 1A) to measure the tip-substrate thermoelectric voltage induced by ΔT. The tip is then slowly withdrawn to a sufficiently large distance (∼15 nm) and the output voltage ΔV is continuously monitored with the tip grounded.

Fig. 1.

Experimental setup and measurements. (A) Schematic description of the experimental set up based on an STM break junction. Molecules of BDT, DBDT, or TBDT are trapped between the Au STM tip kept at ambient temperature and a heated Au substrate kept at temperature ΔT above the ambient. When the tip approaches the substrate, a voltage bias is applied and the current is monitored to estimate the conductance. When the conductance reaches a threshold of 0.1Go, the voltage bias and the current amplifier are disconnected. A voltage amplifier is then used to measure the induced thermoelectric voltage, ΔV, and the tip is gradually pulled away from the substrate. (B) A plot of the thermoelectric voltage measured as a function of the tip-sample distance when a temperature differential ΔT = 20 K is applied (Au tip at ambient and substrate at ambient + 20 K). The blue curve is obtained when a Au-BDT-Au junction is broken. The red curve shows a control experiment performed on a clean gold substrate. (C) Typical thermoelectric voltage traces for tip-substrate temperature differentials of 0, 10, 20, and 30 K for Au-BDT-Au junctions.

When the Au substrate is covered by thiol-terminated molecules, we have shown that molecular bridges are formed between the Au tip and the substrate, and the electrical conductance of single molecules can be monitored in air (24). In our experiment, we covered the Au substrate with BDT, dibezenedithiol (DBDT), or tribenzenedithiol (TBDT) (25) molecules. If molecules of BDT, DBDT, or TBDT are trapped between the tip and substrate with a superimposed ΔT, we should expect to see a thermoelectric voltage generated between the electrodes (8), which should last as long as one or more molecules are trapped and vanish once all of the molecules break away.

A typical thermoelectric voltage curve obtained in the experiment that was performed with a ΔT of 20 K and with the substrate covered with BDT molecules is shown in Fig. 1B. We observed a constant thermoelectric voltage of about ΔV = –200 μV (Fig. 1B, blue curve), which lasted until all of the molecules trapped in the junction broke away. Notably, the distance the tip travels (∼1 to 2 nm) before all of the molecules break away is much longer than the molecular length. Because the thiol group on the molecule binds sufficiently strongly to Au, and because Au atoms are sufficiently mobile at room temperature, it has been proposed that Au chains are formed both on the tip and the substrate when the STM tip is pulled away (6). In contrast to electrical conductance measurements (24), which decrease in steps as molecules break away one at a time, no steps were observed in the ΔV, suggesting that Sjunction is independent of the number of molecules. As the ΔT increases from 0 to 30 K, the thermoelectric voltage signal increases (Fig. 1C). Control experiments performed on clean gold surfaces without any molecules (red, Fig. 1B) demonstrate that no measurable ΔV is generated in the absence of molecules.

To obtain a statistically significant value of ΔV of a Au-BDT-Au junction, we performed roughly 1000 consecutive experiments at each value of ΔT. These data were used to construct histograms for each temperature differential without any data preselection. The histograms thus obtained (Fig. 2, A to C) were used to estimate the average and the variation in Sjunction. The relation between Sjunction of the Au-molecule-Au junction and the measured voltage is (23) Embedded Image(2) where SAu is the Seebeck coefficient of bulk Au, which is ∼1.94 μV/Kat300 K (26). In Fig. 2D, ΔVpeak is plotted as a function of ΔT, where ΔVpeak corresponds to the ΔV at the peak of the distribution. The error bars in Fig. 2D correspond to the full-width half-maximum (FWHM) of the distributions. From the slope ΔVpeakT and Eq. 2, one obtains SAu-BDT-Au = +8.7 ± 2.1 μV/K, where the error is FWHM. Similar experiments were also performed with DBDT and TBDT, and statistical analysis revealed that SAu-DBDT-Au = +12.9 ± 2.2 μV/K and SAu-TBDT-Au =+14.2±3.2 μV/K (23) (Fig. 2E). There seems to be a linear dependence of thermopower with molecular length, which is in contrast to the exponential dependence of electrical resistance that is generally attributed to tunneling across the molecule. The histograms of ΔV for Au-TBDT-Au junctions at 20 and 30 K (fig. S4, A to C) exhibit deviations from a Gaussian curve. Furthermore, the FWHMs for the histograms at 20 and 30 K increased considerably compared with those of Au-BDT-Au junctions, which may arise from the effect of the conformational changes in the molecules trapped in the junctions. The plot of the peak values of the histograms versus the temperature differential for DBDT and TBDT (figs. S3D and S4D) show deviations from linearity. This deviation may arise because the applied temperature differentials of 20 and 30 K across molecules are so high that linear transport theory may not adequately describe the temperature dependence of the thermoelectric voltage.

Fig. 2.

Histograms obtained by analyzing approximately 1000 consecutive thermoelectric voltage curves obtained in measurements of Au-BDT-Au junctions with tip-substrate temperature differential (A) ΔT = 10 K, (B) ΔT = 20 K, and (C) ΔT = 30 K. a.u., arbitrary units. (D) Plot of the peak values of the thermoelectric voltage in histograms as a function of the temperature differential. The error bars represent FWHM of the corresponding histograms. It can be seen that the measured voltage varies linearly with the temperature differential, as expected. (E) Plot of measured junction Seebeck coefficient as a function of molecular length for BDT, DBDT, and TBDT.

The relative position of the HOMO and LUMO levels with respect to the EF of the metal electrodes can be related to the measured value of Sjunction (8). The Landauer formula (27) is used to relate Sjunction to the transmission function, τ(E). It is shown that Sjunction can be obtained as Embedded Image(3) where kB is the Boltzmann constant.

The transmission function for the case of Au-BDT-Au junction, which was derived with the use of the nonequilibrium Green's function formalism in conjunction with extended Huckel theory (8), is shown in Fig. 3A. It is clear that τ(E) ∼1 when the EF aligns with either the HOMO or the LUMO levels and decreases rapidly to below 0.01 in between. Using this transmission function in Eq. 3, we calculated SAu-BDT-Au (Fig. 3B) and found that SAu-BDT-Au is positive (p-type) if EF is closer to the HOMO level and negative (n-type) if it is closer to the LUMO level. Using the measured value of SAu-BDT-Au = +8.7 ± 2.1 μV/K, we see from Fig. 3B that EF is ∼1.2 eV from the HOMO level. The value of the transmission function at this relative position of the Fermi level was τ(E) ∼0.01 (Fig. 3A). In the Landauer formalism, we know that the conductance Gmolecule can be related to the transmission function at EF as Embedded Image(4)

Fig. 3.

Relating the measured Seebeck coefficient of Au-BDT-Au junction to the position of Fermi level. (A) Theoretical prediction (8) of the transmission function of a Au-BDT-Au junction plotted as a function of the relative position of the Fermi level of the Au electrodes with respect to the HOMO and LUMO levels. (B) The predicted (8) Seebeck coefficient of a Au-BDT-Au junction as a function of the relative position of the Fermi level with respect to the HOMO and LUMO levels. When the measured value of SAu-BDT-Au = +8.7 ± 2.1 μV/K (blue band) is shown, it is clear that the Fermi level is ∼1.2 eV above the HOMO level. At this energy level, the transmission function is τ(E) ∼0.01.

Equation 4 implies that the conductance of BDT should be ∼0.01Go. This estimated value of the electrical conductance is in excellent agreement with the measured electrical conductance of Au-BDT-Au junction (11, 28).

Junction Seebeck coefficient measurements can provide insight into the electronic structure of the heterojunction, but the results also bear on an as-yet unexplored field of thermoelectric energy conversion based on molecules. The best efficiency in thermoelectric energy conversion can be achieved if charge transport occurs through a single energy level (29, 30). Single-level transport is, however, difficult to realize in inorganic materials. Metal-molecule-metal heterojunctions are ideal in this regard because they (i) provide transport either through the HOMO or LUMO levels and (ii) have very low vibrational heat conductance because of large mismatch of vibrational spectra between the bulk metal and discrete molecules (31). Hence, such a hybrid material offers the promise of efficient thermoelectric energy conversion. We show for the first time values for molecular junction Seebeck coefficients, but the tunability of this effect and electrical conductance remains unknown. The length dependence of molecular junction Seebeck coefficients is shown in Fig. 2E for the molecules we studied, but there may be other ways of tuning thermopower, such as by introducing various chemical moieties in the molecule or by controlling the metal-molecule chemical bond.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1137149/DC1

Materials and Methods

Figs. S1 to S6

References

References and Notes

View Abstract

Navigate This Article